The Currency You Didn’t Choose

May 10, 2026

Run the same three-asset strategy out of New York and out of Frankfurt. The American gets Sharpe 0.97 and a 22% drawdown. The European, holding identical positions but spending in euros, gets Sharpe 0.65 and a 45% drawdown. The trades are the same. The difference is a currency position the European never chose to take, sized by the strategy’s gross exposure rather than by any view on EUR/USD.

What the recent data shows

Before getting to the strategy, look at what most retail portfolios actually hold: MSCI World, the broad developed-market equity index, roughly seventy percent US large-cap and thirty percent developed-international. Compounded since January 2009 in two currency views — what a US investor experienced in dollars, and what an unhedged euro investor experienced after translation.

Figure 1 — MSCI World as a 70% VTI / 30% VEA proxy, monthly rebalanced, 2009 to today. The unhedged euro investor outperformed the US investor by 130 basis points per year. Drawdowns are nearly identical at around minus thirty-four percent for both. From this seventeen-year sample alone, the case for hedging USD exposure is hard to make.

CAGR 12.1% for the US investor, 13.4% for the unhedged euro investor. Sharpe 0.79 versus 0.72. Drawdowns essentially identical at minus thirty-four percent. The 2020 COVID drawdown looks worse than 2022. The 2022 simultaneous stock-bond crash looks like a normal bear market. There is no visible currency-related disaster anywhere on this chart.

The same lesson holds for a more sophisticated strategy. An inverse-volatility-weighted three-asset portfolio of US equity, long Treasuries, and gold, with monthly rebalancing, an 8% volatility target, an 8.5% trigger band, and no leverage above one — the IVOL strategy I have written about before in Beyond 60/40, Volatility Targeting, and Risk Premia in Five Decades. A clean barbell of three orthogonal risk premia. The full-sample numbers will come back into the picture in a moment.

Same post-2008 sample, USD versus unhedged EUR:

Figure 2 — The IVOL strategy 2009 to today, US investor versus unhedged euro investor. The euro investor outperformed by over 100 basis points per year and had a smaller drawdown — 14.0% versus 16.7%. The 2022 simultaneous stock-bond crash, the worst event in this period for the US investor, was less severe for the unhedged euro investor because the dollar strengthened during the stress.

CAGR 9.31% for the unhedged euro investor versus 8.22% for the US investor. Sharpe 0.93 versus 1.07 — the euro investor took on more volatility but earned more return for it. Maximum drawdown 14.0% versus 16.7% — the euro investor actually had a shallower bottom. The 2022 stock-bond stress, which was the worst moment in this period for the US investor, was comparatively mild for the unhedged euro investor because the dollar strengthened sharply during that stress. The currency leg worked as a hedge exactly when the strategy needed one.

Two charts, two strategies, one apparent message: hedging USD exposure has not been worth doing in the recent past. The unhedged euro investor outperformed in every metric that matters. There is no visible cost on either picture. Why hedge?

What the recent data hides

The answer becomes obvious only when the same comparison runs over the full sample. Here is the IVOL strategy from 1970, US versus unhedged EUR, with the drawdown panel below.

Figure 3 — The same comparison extended back to 1970. The US investor’s CAGR of 7.08% and Sharpe of 0.95 collapse to 5.58% CAGR and Sharpe 0.65 for the unhedged euro investor. The maximum drawdown roughly doubles from minus 23.0% to minus 44.9%. The deepest unhedged drawdowns happened in 1973–1980 (post-Bretton Woods USD slide) and 1985–1995 (post-Plaza dollar collapse) — episodes invisible in the post-2008 sample. The 2022 event, which barely registered for the unhedged investor in Figure 2, lives at the right edge of a much longer history of currency damage.

The numbers reset. CAGR 7.08% for the US investor, 5.58% for the unhedged euro investor. Sharpe 0.95 versus 0.65. Maximum drawdown minus 23.0% versus minus 44.9% — the unhedged investor’s drawdown is essentially double the US investor’s. The drawdown panel makes the regime structure unmistakable. Two long episodes of severe FX damage sit in the 1970s and the 1985–1995 stretch — the post-Bretton-Woods dollar slide and the post-Plaza-Accord dollar collapse, respectively. Neither shows up in Figure 1 or Figure 2 because neither happened after 2008.

For an investor who has been paying attention only to the post-2008 sample, the right edge of Figure 3 looks normal. For an investor who started thinking about portfolios in 1973 or 1985, what shows up in Figures 1 and 2 looks like a remarkable interruption of normal currency behavior. Both readings are correct on their own samples. The honest reading is that the recent regime is one regime among several, and the historical record contains FX episodes that are an order of magnitude worse than anything visible in the post-2008 sample.

For a saver compounding into the strategy with monthly contributions, or a retiree drawing on it, the difference between a 23% and a 45% maximum drawdown is not a small detail. Sequence-of-returns risk does not care about CAGR. It cares about the specific moment a deep drawdown lands on a portfolio that is being drawn from. The forty-five-percent unhedged drawdown is a real risk of ruin for an investor in or near decumulation, and the post-2008 picture provides no warning that such a drawdown is even possible.

Where the carry comes from

The natural next question is what the actual cost of hedging would have been. The hedge cost is set by the rate differential between USD and EUR short rates, applied to whatever notional the strategy is carrying. Here is the full history of that differential, with US Fed Funds and the European short rate (Bundesbank Diskontsatz pre-1999, ECB Deposit Facility Rate post-1999) plotted together.

Figure 4 — Short-rate levels and the US–EUR differential, 1970–2026. Top: US Fed Funds versus the European short rate, with the gap shaded by sign. Bottom: the differential itself. The carry cost a hedger pays equals this differential applied to the strategy’s gross exposure. The Volcker era 1979–1985 had positive differentials of ten to fifteen percent — the hedger paid heavily during those years. The 1992 Bundesbank tightening flipped the sign and paid the hedger five percent on notional. Post-2008 the differential has hovered near zero or slightly positive. Pre-1999 EUR rate is the Bundesbank Diskontsatz, an administered floor rate; market rates ran roughly fifty to one hundred fifty basis points above this number.

The differential is large and time-varying. It has spent significant time at plus ten percent and a smaller stretch at minus five percent. A static hedge collects whatever this number happens to be in real time. The Volcker disinflation, the post-Plaza collapse, the German reunification tightening, and the post-2014 Fed–ECB divergence are all visible directly. None of this was avoidable for a hedged euro investor; they paid or received whatever the differential delivered, applied to their notional.

The implicit position

Every euro the European investor allocates to a dollar-denominated asset is, in foreign-exchange terms, a short EUR/USD position. When the euro strengthens against the dollar, the dollar holdings translate to fewer euros and the investor loses on the FX leg. When the dollar strengthens, they gain. The size of this short is mechanical: it equals the strategy’s gross exposure on any given day. Vol-target downscaling reduces it, leverage caps it at one. The median is one hundred percent and the mean is eighty-nine percent — most of the time the European investor is carrying a full-notional short EUR/USD without thinking about it.

The cumulative P&L of that implicit position, isolated from the asset returns, is shown below.

Figure 5 — The implicit short EUR/USD position carried by an unhedged EUR investor. Top: cumulative FX-leg equity, falling from 1.0 in 1970 to 0.44 by 2026. Middle: notional in percent of NAV. Bottom: rolling 24-month annualized volatility of the strategy versus the FX leg. The euro investor is running two roughly equal risks stacked, with the one they did not choose contributing nothing to expected return on average over the full sample.

The cumulative equity of the position is 0.44× over the full sample. The investor lost fifty-six percent of unit notional on the FX leg over fifty-six years — or rather, would have, had this been a standalone position the investor sized themselves. The volatility of the FX leg has been comparable to the volatility of the strategy itself in the modern era — both run around seven to nine percent annualized in the post-1999 sample. The European investor is therefore not running one risk; they are running two roughly equal risks stacked.

What the position actually does

The next question is whether the FX leg has any redeeming property at all. The honest answer is that it is two separate things. There is a tail-protective component, which appears specifically when the asset book is hurting, and there is a directional drift component, which dominates the cumulative outcome over multi-year windows.

Figure 6 — Decomposition of the FX-leg return. Top: cumulative FX-leg equity with IVOL bottom-5% months marked. Bottom: per-decade decomposition. Green bars show the FX return earned during IVOL’s worst-five-percent months — the tail-conditional component. Navy bars show the FX return during the other ninety-five percent of months — the directional drift. The red diamond is the total decade outcome. Tail components are small and mostly positive across decades; directional components are large and bipolar.

The 2000s show what is going on most clearly. In that decade, the worst-five-percent months for the IVOL strategy delivered a cumulative FX return of plus twenty-five percent — the position made money exactly when the asset book was under stress, every time. And the other ninety-five percent of months delivered a cumulative return of minus forty-four percent, pushing the decade total to minus twenty-nine. The position acted as a partial hedge during the dot-com unwind and the Global Financial Crisis, and then bled steadily across the years between, which were dominated by a multi-year dollar-weakening trend tied to the post-2000 USD slide and the global savings glut that financed it.

The 1970s tell the opposite story without the tail-protective property. After Bretton Woods broke and the dollar entered a long secular decline, the FX leg lost almost half of unit notional over the decade and produced no measurable tail benefit. The 1980s had no tail content either, and the post-Plaza dollar collapse made the directional component the main driver of the decade’s loss. The 2010s and 2020s brought the tail behavior back in stronger form — five of six and four of four IVOL bottom-five-percent months had positive FX returns — and the directional component finally began to add modestly to wealth as the dollar strengthened structurally. This is the regime that produces Figures 1 and 2.

Per-decade detail

Figure 7 — Per-decade tail analysis. Each row is one decade plus the full sample. Left panels are return distributions; right panels are scatter and OLS regression of FX-leg monthly return on IVOL monthly return. Red dots are IVOL bottom-five-percent months. The regression beta varies from zero in the 1970s to minus 0.555 in the 2000s, with R² ranging from zero to thirty-four percent. The post-2008 decades show the tightest negative beta and the most consistent tail behavior; the pre-1999 decades have artificially compressed FX volatility because the underlying daily path is interpolated.

The slope of the regression line — how much the FX leg moves for a one-percent move in the IVOL strategy — varies from zero in the 1970s to minus 0.39 in the 2020s, passing through minus 0.56 in the 2000s. The R² varies from less than one percent to thirty-four percent. The tendency of the FX leg to systematically work in the asset book’s favor at the worst moments is not a constant of nature; it is a property of the post-1999 international monetary regime, and it is strongest exactly where central-bank coordination and global flight-to-USD flows have been most pronounced.

The two extremes

The two simple policies. The unhedged investor takes the cumulative FX P&L of the implicit short — the red curve in the figure below. The fully hedged investor sells dollars forward each period at the prevailing forward rate and pays the realized interest differential as a deterministic cost — the green curve. To verify the math from a different direction, the navy dashed curve shows the realized rolling cost of executing the same hedge through CME EUR/USD futures, where the rollover between contracts embeds the carry that arbitrage-free pricing requires.

Over this fifty-six-year sample the two extremes happen to land at similar cumulative wealth. That is one path realization, not a guarantee. Covered interest parity holds in expectation, not on every realized path. The unhedged investor pays the cost in random FX swings — they could have arrived at any point along a wide distribution. The hedged investor pays it in a smooth, deterministic carry drag — their distribution is much narrower around the realized rate differential.

The volatility profiles are very different. The unhedged FX P&L runs at 5.8% annualized vol with a peak-to-trough drawdown of seventy percent. The hedge carry cost runs at 0.13% vol with a maximum drawdown of about ten percent, concentrated in the early 1990s when the Bundesbank held rates above the Fed during the German reunification period. The hedger has converted a noisy seven-percent-vol risk into a near-deterministic stream of payments. The unhedged investor still carries that seven-percent-vol risk on top of the asset book — and Figure 6 shows that this risk includes a partial tail-protective property which a full hedge throws away.

These are the two extremes that bracket every other choice. Any partial hedge, any time-varying hedge, any conditional rule sits somewhere in between. The unhedged path embeds tail behavior that has historically helped during the worst months for the asset book, but at the cost of a large drawdown distribution that matters enormously for any investor saving into or drawing from the portfolio. The fully hedged path removes both the drawdown problem and the tail benefit, and pays a deterministic interest differential for the privilege.

What we have

The post-2008 sample is unusual. It contains one of the strongest USD bull markets in fifty years and one of the longest equity bull markets in history, stacked. For an investor whose mental model of “currency risk” was formed in this window, FX hedging genuinely looks unnecessary. From the data they have seen, it is.

The full sample tells a different story. Carrying USD exposure unmanaged adds a forty-five-percent maximum drawdown to a strategy that would otherwise have drawn down twenty-two percent — twice the worst case. For a saver compounding into the portfolio or a retiree drawing from it, that doubling is not a small detail. The deepest historical FX episodes happened before the typical retail investor’s memory begins.

The implicit short EUR/USD position has a useful property — it has worked as a partial tail hedge when the asset book is hurting — but it also has a directional drift component that is large, bipolar across decades, and has been costly more often than not. The two simple answers, never hedge and always hedge, both leave money on the table in different ways. Neither extreme is satisfactory.

The next article asks whether information available at decision time — rate differentials, FX trend, regime indicators — can be used to size the hedge dynamically and capture the tail-protective property of being unhedged without paying its full directional cost.

Notes & references

The two trivial-hedge methods — applying the realized USD–EUR rate differential daily versus the realized roll cost of CME 6E EUR/USD futures — agree on the post-1999 sample to within 0.13 percentage points per year. They would agree exactly under covered interest parity. The small remaining gap is consistent with the documented post-2008 cross-currency basis, the dislocation between FX-swap-implied USD funding and direct USD funding that emerged after the global financial crisis. See Borio, McCauley, McGuire, and Sushko (2016), “Covered interest parity lost: understanding the cross-currency basis,” BIS Quarterly Review, September; and Du, Tepper, and Verdelhan (2018), “Deviations from Covered Interest Rate Parity,” Journal of Finance 73(3), 915–957 (NBER WP 23170). The basis is real and well-documented but does not affect the qualitative argument of this article.
Pre-1999 EUR/USD is reconstructed from monthly DM/USD anchors sourced from public economic-history references (BIS, IMF IFS) using the irrevocable conversion rate of 1 EUR = 1.95583 DM. Daily paths between monthly anchors are linearly interpolated. Pre-1999 numbers are accurate at month boundaries; daily-frequency volatility before that date is artificially smoothed and should not be read off Figure 5 or Figure 7 panels for the 1970s and 1990s.

Discussion about this post

Ready for more?